Most travel planning is scattered across emails, booking sites, maps and notes; TREK brings those pieces into a single, self-hosted app and lets AI assistants act on the canonical itinerary. The pragmatic angle: real-time collaboration + PWA/offline sync for people who want control over their travel data, plus an embedded MCP server so agents can read and modify trips with fine-grained scopes.
What Sets It Apart
- Self-hosted, collaborative trip management with real-time WebSocket sync and PWA support — usable offline and installable on mobile without app stores. This makes it suitable for travel groups who want a shared, always-consistent plan.
- Built-in MCP (Model Context Protocol) server with OAuth 2.1 and granular scopes. AI clients can be authorized to read trips, places, budgets and more, and TREK provides pre-built prompts (e.g. trip-summary, packing-list, budget-overview) so assistants get structured context without copy/paste.
- Practical travel features beyond AI: interactive maps (Leaflet or Mapbox GL), route optimisation, reservations import (email/PDF/KDE Itinerary), multi-currency expense splitting, packing lists and document attachments, PDF export — all manageable from an admin panel and backed by SQLite for single-node deployments.
Who It's For and Trade-offs
Great fit if you want a self-hosted travel hub that integrates with AI assistants: small teams, privacy-conscious travelers, travel clubs or companies that prefer data residency. It’s also convenient for users who want offline-capable PWAs and real-time multi-user editing. Look elsewhere if you need a SaaS-scale, multi-datacenter system out of the box: TREK is designed for single-instance self-hosting (SQLite by default) and exposes WebSockets (requires a reverse proxy that supports upgrades). The project is AGPL-3.0 — offering a modified network service requires publishing your changes. Some integrations (optional place searches, third-party tiles) depend on external APIs or API keys.
